基于不同光照方向的子空间人脸识别方法  

Face Recognition Based on Subspace of Variant Lighting Direction

作　　者：苏宏涛[1] 赵荣椿[1] 

机构地区：[1]西北工业大学计算机科学与工程系,西安710072

出　　处：《计算机工程》2003年第8期124-125,共2页Computer Engineering

摘　　要：提出了一种方法，用于解决人脸识别问题中光照方向变化所带来的困难。首先建立对应不同光照方向的子空间，再确定测试样本所对应的子空间，然后在这个子空间中利用最近邻法进行匹配，达到识别的目的。利用SHIMON人脸数据库，分仅有光照变化和同时有视角变化和光照变化两种情况进行了实验，实验结果证明该方法是有效的。This paper presents a method to solve the problem caused by the variant lighting direction in face recognition.The subspaces corresponding to different lighting direction are constructed, and the subspace that the test sample belongs to is determined, then a match using nearest neighboring method is carried out and the ultimate recognition result is achieved. The experiment is divided into two cases based on SHIMON face database, there is only lighting direction changing in the first case, in another case there are both lighting direction changing and view changing. The experiment result proves that this method is effective.

关 键 词：人脸识别 子空间 本征脸 光照方向 视角变化 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]